kvm: require KVM_CAP_IOEVENTFD and KVM_CAP_IOEVENTFD_ANY_LENGTH

KVM_CAP_IOEVENTFD_ANY_LENGTH was added in Linux 4.4, released in 2016.
Assume that it is present.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

accel/kvm/kvm-all.c

@@ -90,7 +90,6 @@ bool kvm_kernel_irqchip;
 bool kvm_split_irqchip;
 bool kvm_async_interrupts_allowed;
 bool kvm_halt_in_kernel_allowed;
-bool kvm_eventfds_allowed;
 bool kvm_resamplefds_allowed;
 bool kvm_msi_via_irqfd_allowed;
 bool kvm_gsi_routing_allowed;
@@ -98,7 +97,6 @@ bool kvm_gsi_direct_mapping;
 bool kvm_allowed;
 bool kvm_readonly_mem_allowed;
 bool kvm_vm_attributes_allowed;
-bool kvm_ioeventfd_any_length_allowed;
 bool kvm_msi_use_devid;
 bool kvm_has_guest_debug;
 static int kvm_sstep_flags;
@@ -110,6 +108,8 @@ static const KVMCapabilityInfo kvm_required_capabilites[] = {
     KVM_CAP_INFO(DESTROY_MEMORY_REGION_WORKS),
     KVM_CAP_INFO(JOIN_MEMORY_REGIONS_WORKS),
     KVM_CAP_INFO(INTERNAL_ERROR_DATA),
+    KVM_CAP_INFO(IOEVENTFD),
+    KVM_CAP_INFO(IOEVENTFD_ANY_LENGTH),
     KVM_CAP_LAST_INFO
 };
 
@@ -2547,18 +2547,12 @@ static int kvm_init(MachineState *ms)
     kvm_readonly_mem_allowed =
         (kvm_check_extension(s, KVM_CAP_READONLY_MEM) > 0);
 
-    kvm_eventfds_allowed =
-        (kvm_check_extension(s, KVM_CAP_IOEVENTFD) > 0);
-
     kvm_resamplefds_allowed =
         (kvm_check_extension(s, KVM_CAP_IRQFD_RESAMPLE) > 0);
 
     kvm_vm_attributes_allowed =
         (kvm_check_extension(s, KVM_CAP_VM_ATTRIBUTES) > 0);
 
-    kvm_ioeventfd_any_length_allowed =
-        (kvm_check_extension(s, KVM_CAP_IOEVENTFD_ANY_LENGTH) > 0);
-
 #ifdef KVM_CAP_SET_GUEST_DEBUG
     kvm_has_guest_debug =
         (kvm_check_extension(s, KVM_CAP_SET_GUEST_DEBUG) > 0);
@@ -2595,19 +2589,15 @@ static int kvm_init(MachineState *ms)
         kvm_irqchip_create(s);
     }
 
-    if (kvm_eventfds_allowed) {
+    s->memory_listener.listener.eventfd_add = kvm_mem_ioeventfd_add;
-        s->memory_listener.listener.eventfd_add = kvm_mem_ioeventfd_add;
+    s->memory_listener.listener.eventfd_del = kvm_mem_ioeventfd_del;
-        s->memory_listener.listener.eventfd_del = kvm_mem_ioeventfd_del;
-    }
     s->memory_listener.listener.coalesced_io_add = kvm_coalesce_mmio_region;
     s->memory_listener.listener.coalesced_io_del = kvm_uncoalesce_mmio_region;
 
     kvm_memory_listener_register(s, &s->memory_listener,
                                  &address_space_memory, 0, "kvm-memory");
-    if (kvm_eventfds_allowed) {
+    memory_listener_register(&kvm_io_listener,
-        memory_listener_register(&kvm_io_listener,
+                             &address_space_io);
-                                 &address_space_io);
-    }
     memory_listener_register(&kvm_coalesced_pio_listener,
                              &address_space_io);
 

accel/stubs/kvm-stub.c

@@ -17,14 +17,12 @@
 KVMState *kvm_state;
 bool kvm_kernel_irqchip;
 bool kvm_async_interrupts_allowed;
-bool kvm_eventfds_allowed;
 bool kvm_resamplefds_allowed;
 bool kvm_msi_via_irqfd_allowed;
 bool kvm_gsi_routing_allowed;
 bool kvm_gsi_direct_mapping;
 bool kvm_allowed;
 bool kvm_readonly_mem_allowed;
-bool kvm_ioeventfd_any_length_allowed;
 bool kvm_msi_use_devid;
 
 void kvm_flush_coalesced_mmio_buffer(void)

hw/misc/pci-testdev.c

@@ -245,7 +245,6 @@ static void pci_testdev_realize(PCIDevice *pci_dev, Error **errp)
     uint8_t *pci_conf;
     char *name;
     int r, i;
-    bool fastmmio = kvm_ioeventfd_any_length_enabled();
 
     pci_conf = pci_dev->config;
 
@@ -279,7 +278,7 @@ static void pci_testdev_realize(PCIDevice *pci_dev, Error **errp)
         g_free(name);
         test->hdr->offset = cpu_to_le32(IOTEST_SIZE(i) + i * IOTEST_ACCESS_WIDTH);
         test->match_data = strcmp(IOTEST_TEST(i), "wildcard-eventfd");
-        if (fastmmio && IOTEST_IS_MEM(i) && !test->match_data) {
+        if (IOTEST_IS_MEM(i) && !test->match_data) {
             test->size = 0;
         } else {
             test->size = IOTEST_ACCESS_WIDTH;

hw/s390x/virtio-ccw.c

@@ -768,10 +768,6 @@ static void virtio_ccw_device_realize(VirtioCcwDevice *dev, Error **errp)
         sch->cssid, sch->ssid, sch->schid, sch->devno,
         ccw_dev->devno.valid ? "user-configured" : "auto-configured");
 
-    if (kvm_enabled() && !kvm_eventfds_enabled()) {
-        dev->flags &= ~VIRTIO_CCW_FLAG_USE_IOEVENTFD;
-    }
-
     /* fd-based ioevents can't be synchronized in record/replay */
     if (replay_mode != REPLAY_MODE_NONE) {
         dev->flags &= ~VIRTIO_CCW_FLAG_USE_IOEVENTFD;

hw/virtio/vhost-user.c

@@ -264,11 +264,6 @@ struct scrub_regions {
     int fd_idx;
 };
 
-static bool ioeventfd_enabled(void)
-{
-    return !kvm_enabled() || kvm_eventfds_enabled();
-}
-
 static int vhost_user_read_header(struct vhost_dev *dev, VhostUserMsg *msg)
 {
     struct vhost_user *u = dev->opaque;
@@ -1318,7 +1313,7 @@ static int vhost_set_vring_file(struct vhost_dev *dev,
         .hdr.size = sizeof(msg.payload.u64),
     };
 
-    if (ioeventfd_enabled() && file->fd > 0) {
+    if (file->fd > 0) {
         fds[fd_num++] = file->fd;
     } else {
         msg.payload.u64 |= VHOST_USER_VRING_NOFD_MASK;

hw/virtio/virtio-mmio.c

@@ -761,10 +761,6 @@ static void virtio_mmio_realizefn(DeviceState *d, Error **errp)
     qbus_init(&proxy->bus, sizeof(proxy->bus), TYPE_VIRTIO_MMIO_BUS, d, NULL);
     sysbus_init_irq(sbd, &proxy->irq);
 
-    if (!kvm_eventfds_enabled()) {
-        proxy->flags &= ~VIRTIO_IOMMIO_FLAG_USE_IOEVENTFD;
-    }
-
     /* fd-based ioevents can't be synchronized in record/replay */
     if (replay_mode != REPLAY_MODE_NONE) {
         proxy->flags &= ~VIRTIO_IOMMIO_FLAG_USE_IOEVENTFD;

hw/virtio/virtio-pci.c

@@ -332,7 +332,6 @@ static int virtio_pci_ioeventfd_assign(DeviceState *d, EventNotifier *notifier,
     VirtQueue *vq = virtio_get_queue(vdev, n);
     bool legacy = virtio_pci_legacy(proxy);
     bool modern = virtio_pci_modern(proxy);
-    bool fast_mmio = kvm_ioeventfd_any_length_enabled();
     bool modern_pio = proxy->flags & VIRTIO_PCI_FLAG_MODERN_PIO_NOTIFY;
     MemoryRegion *modern_mr = &proxy->notify.mr;
     MemoryRegion *modern_notify_mr = &proxy->notify_pio.mr;
@@ -343,13 +342,8 @@ static int virtio_pci_ioeventfd_assign(DeviceState *d, EventNotifier *notifier,
 
     if (assign) {
         if (modern) {
-            if (fast_mmio) {
+            memory_region_add_eventfd(modern_mr, modern_addr, 0,
-                memory_region_add_eventfd(modern_mr, modern_addr, 0,
+                                      false, n, notifier);
-                                          false, n, notifier);
-            } else {
-                memory_region_add_eventfd(modern_mr, modern_addr, 2,
-                                          false, n, notifier);
-            }
             if (modern_pio) {
                 memory_region_add_eventfd(modern_notify_mr, 0, 2,
                                               true, n, notifier);
@@ -361,13 +355,8 @@ static int virtio_pci_ioeventfd_assign(DeviceState *d, EventNotifier *notifier,
         }
     } else {
         if (modern) {
-            if (fast_mmio) {
+            memory_region_del_eventfd(modern_mr, modern_addr, 0,
-                memory_region_del_eventfd(modern_mr, modern_addr, 0,
+                                      false, n, notifier);
-                                          false, n, notifier);
-            } else {
-                memory_region_del_eventfd(modern_mr, modern_addr, 2,
-                                          false, n, notifier);
-            }
             if (modern_pio) {
                 memory_region_del_eventfd(modern_notify_mr, 0, 2,
                                           true, n, notifier);

include/sysemu/kvm.h

@@ -36,13 +36,11 @@ extern bool kvm_kernel_irqchip;
 extern bool kvm_split_irqchip;
 extern bool kvm_async_interrupts_allowed;
 extern bool kvm_halt_in_kernel_allowed;
-extern bool kvm_eventfds_allowed;
 extern bool kvm_resamplefds_allowed;
 extern bool kvm_msi_via_irqfd_allowed;
 extern bool kvm_gsi_routing_allowed;
 extern bool kvm_gsi_direct_mapping;
 extern bool kvm_readonly_mem_allowed;
-extern bool kvm_ioeventfd_any_length_allowed;
 extern bool kvm_msi_use_devid;
 
 #define kvm_enabled()           (kvm_allowed)
@@ -86,15 +84,6 @@ extern bool kvm_msi_use_devid;
  */
 #define kvm_halt_in_kernel() (kvm_halt_in_kernel_allowed)
 
-/**
- * kvm_eventfds_enabled:
- *
- * Returns: true if we can use eventfds to receive notifications
- * from a KVM CPU (ie the kernel supports eventds and we are running
- * with a configuration where it is meaningful to use them).
- */
-#define kvm_eventfds_enabled() (kvm_eventfds_allowed)
-
 /**
  * kvm_irqfds_enabled:
  *
@@ -147,12 +136,6 @@ extern bool kvm_msi_use_devid;
  */
 #define kvm_readonly_mem_enabled() (kvm_readonly_mem_allowed)
 
-/**
- * kvm_ioeventfd_any_length_enabled:
- * Returns: true if KVM allows any length io eventfd.
- */
-#define kvm_ioeventfd_any_length_enabled() (kvm_ioeventfd_any_length_allowed)
-
 /**
  * kvm_msi_devid_required:
  * Returns: true if KVM requires a device id to be provided while
@@ -168,13 +151,11 @@ extern bool kvm_msi_use_devid;
 #define kvm_async_interrupts_enabled() (false)
 #define kvm_halt_in_kernel() (false)
 #define kvm_irqfds_enabled() (false)
-#define kvm_eventfds_enabled() (false)
 #define kvm_resamplefds_enabled() (false)
 #define kvm_msi_via_irqfd_enabled() (false)
 #define kvm_gsi_routing_allowed() (false)
 #define kvm_gsi_direct_mapping() (false)
 #define kvm_readonly_mem_enabled() (false)
-#define kvm_ioeventfd_any_length_enabled() (false)
 #define kvm_msi_devid_required() (false)
 
 #endif  /* CONFIG_KVM_IS_POSSIBLE */

system/memory.c

@@ -1535,7 +1535,12 @@ MemTxResult memory_region_dispatch_write(MemoryRegion *mr,
 
     adjust_endianness(mr, &data, op);
 
-    if ((!kvm_eventfds_enabled()) &&
+    /*
+     * FIXME: it's not clear why under KVM the write would be processed
+     * directly, instead of going through eventfd.  This probably should
+     * test "tcg_enabled() || qtest_enabled()", or should just go away.
+     */
+    if (!kvm_enabled() &&
         memory_region_dispatch_write_eventfds(mr, addr, data, size, attrs)) {
         return MEMTX_OK;
     }
@@ -2550,8 +2555,6 @@ void memory_region_clear_flush_coalesced(MemoryRegion *mr)
     }
 }
 
-static bool userspace_eventfd_warning;
-
 void memory_region_add_eventfd(MemoryRegion *mr,
                                hwaddr addr,
                                unsigned size,
@@ -2568,13 +2571,6 @@ void memory_region_add_eventfd(MemoryRegion *mr,
     };
     unsigned i;
 
-    if (kvm_enabled() && (!(kvm_eventfds_enabled() ||
-                            userspace_eventfd_warning))) {
-        userspace_eventfd_warning = true;
-        error_report("Using eventfd without MMIO binding in KVM. "
-                     "Suboptimal performance expected");
-    }
-
     if (size) {
         adjust_endianness(mr, &mrfd.data, size_memop(size) | MO_TE);
     }